Book Club Review: The AI Con

At Engage, a handful of us get together once a week as a book club to read and discuss selected titles over the course of a couple months. We recently read The AI Con by Alex Hanna and Emily M. Bender. As you might expect from a book titled The AI Con: How to Fight Big Tech’s Hype and Create the Future We Want, the outlook is generally critical of the vast majority of AI use cases. Here’s an overview of the book, which may or may not match the specific opinions of anyone at Engage.

What even is AI? It’s an intentionally murky marketing term. Tech companies are trying to convince you that machines can think, but really they’re just using advanced math to predict what words make sense next. They demonstrate how even in the 1960s, a simple program called ELIZA was able to fool even people who knew how it worked, because our brains are wired to see intention behind words. But accepting the belief that there is thinking behind these technologies can distract us from the real harms and dangers that can affect real people in the here-and-now.

The main thrust of the book is that the hype around AI doesn’t deliver and also distracts. The underlying algorithms that Large Language Models are based upon cannot reason or replicate thinking, as much as we may be wired to see thinking behind the words they generate. When AI companies promise to reduce the toil of office workers, what really happens is that workers have to sift through more workslop and that managers layoff coworkers based on that promise, leaving more work to be completed by fewer people. When AI companies promise to expand access to medical and legal advice, what really happens is that bad advice spreads, making more work for the overworked professionals, and reducing the respect and compensation they’re able to retain. When AI companies promise to expand access to creative pursuits, what really happens is that journalists are fired and scientists are inundated with novel ideas that don’t pan out.

While the book spends a good amount of effort debunking the myths around computer intelligence and exposing how AI can be discriminatory, put data at risk, and trade-in human decision making for computers, shifting accountability to “profit-without-blame”, it doesn’t really provide us with the tools we’re looking for to deconstruct the hype. There are some thoughts and answers, but not enough to convince someone who’s not already questioning the place of the tech giants in our lives.

Many of our concerns around how AI is shaping the future goes beyond just tech and everyday tools, and encroaches on societal and political realms that have a lot more to unpack. The threat is not that of AI taking over the world, but of how shifting to large language model outputs for human-centered endeavors like social services, scientific research, creative work, legal decisions, and the like can lead to centralized power, further exploitation of data, and the prioritization of profit instead of striving to create true societal benefit. Large language models are tools, and when utilized should be treated as such. Anthropomorphizing AI gives the wrong impression that AI is human-like, which changes the way we think about how the tools are being used and exactly what they are actually capable of.